Repository navigation
Stop the ISRM tutorial's run_isrm from hanging on a silent worker - #124
Merged
Merged
Conversation
The EarthSciLab event stream sends a keep-alive every 15 s, so watch()'s
240 s socket timeout never fired on a quiet stream and it never fell back to
GET /runs/{id}. When a cloud worker died without reporting back, run_isrm
waited forever. After 60 s with no real event, watch() now checks the run
directly and returns if it has ended.
Same change as EarthSciML/isrm.esm dbf4b12, in both the post's code block and
its notebook copy.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
run_isrmin the ISRM on EarthSciLab tutorial could hang forever. The EarthSciLab event stream sends a keep-alive every 15 s, sowatch()'s 240 s socket timeout never fires on a quiet stream, andwatch()only checksGET /runs/{id}after the stream drops. When a cloud worker died without reporting back, nothing ever ended the wait.watch()now checks the run directly after 60 s with no real event, and returns if the run has ended. The change is identical in the post's code block (website/blog/2026-09-09-isrm-esm.md) and in its notebook copy (website/static/blog/2026-09-09-isrm-esm/isrm_esm.ipynb). It is the same as EarthSciML/isrm.esm@dbf4b12.The EarthSciLab server has also been fixed so the event stream itself reports a dead worker as
failedwithin about 30 s. With that deployed, this change is a second safeguard rather than the only fix.Testing
watch()returnfailedafter one status check, about 75 s in.masteronly by the 17 added lines.🤖 Generated with Claude Code